Making the Earth Greener, One Bite at a Time

Computers are useless. They can only give you answers
-- Pablo Picasso --

At Blue Whale Software, we strongly believe in improving integration with nature not via throwing technology away, but through making technology better. There were already numerous examples when technology first affected environment in negative way (sometimes for many years), but subsequent technology improvements have made environment much better once again.

We know that amounts of power consumed by computers are enormous these days. One single datacenter alone can use as much as 5 Megawatt, and overall in the world there are about 1 billion computers (see Forrester Research press release) which, assuming 200W per computer and that 50% of the time they are turned off, consume massive 100000 Megawatt, or about 1/4th of the worldwide nuclear plant power. It means that by decreasing computer power consumption by 50% it is possible to eliminate 1/8th of the worldwide nuclear reactors or reduce equivalent amounts of fossil fuel emissions. Obviously, these estimates are extremely rough, but they still should give an idea about the scale of the problem.

And reducing CPU power consumption (yes, we know it is not the same as reducing overall computer power consumption) by 50% is not that difficult task, as it might seem. The thing is that CPUs tend to consume much more when they're working in extreme modes (close to top performance). Just as an example: Intel Core 2 Extreme QX9775 (45 nm, February/2008) has TDP of 150W providing 4 cores @ 3.2GHz, and mobile Intel Core 2 Quad Q9100 (45 nm, Aug/2008) has TDP of 45W providing 4 cores @ 2.2Ghz, making it (very roughly) 3x reduction in power in exchange to just 30% decrease in performance. While it would be way too optimistic to assume that improving 30% on performance will always save 3x in power consumption, highly non-linear dependency between CPU speed and power consumption is obvious, making even modest performance optimizations (such as 10-20%) potentially much more attractive from power consumption perspective. In addition, it is fairly obvious that the more memory the software needs, the more power it will consume, so reducing memory consumption (and hard disk space consumption) will also reduce power needed to feed computers.